Motion Clouds: raw principles¶

Motion Clouds are synthesized textures which aim at having similar characteristics as natural images but with controlled parameters. There are many ways to achieve these results and this notebook aims at showing that different procedures from different communities (neurioscience, modelling, computer vision, ...) may produce similar results.

In [6]:

import numpy as np
np.set_printoptions(precision=3, suppress=True)
import pylab
import matplotlib.pyplot as plt
%matplotlib inline

In [7]:

def show_video(filename):
    from IPython.display import display, Image, HTML
    from base64 import b64encode
    video = open(mc.figpath + filename + mc.vext, "rb").read()
    video_encoded = b64encode(video)
    video_tag = '<video controls  autoplay="autoplay" loop="loop" width=350px src="data:video/x-m4v;base64,{0}">'.format(video_encoded)
    display(HTML(data=video_tag))

Using Fourier ("official Motion Clouds")¶

In [8]:

import sys
sys.path.append('..')
import MotionClouds as mc
fx, fy, ft = mc.get_grids(mc.N_X, mc.N_Y, mc.N_frame)

Using mixtures of images¶

In [9]:

import scipy.misc
lena = scipy.misc.lena() * 1.
lena -= lena.mean()
lena /= lena.std()
print lena.shape

(512, 512)

In [11]:

plt.imshow(lena, cmap=plt.cm.gray)

Out[11]:

<matplotlib.image.AxesImage at 0x104fdb450>

In [12]:

def noise(image=lena):
    for axis in [0, 1]:
        image = np.roll(image, np.random.randint(image.shape[axis]), axis=axis)
    return image

In [13]:

plt.imshow(noise(), cmap=plt.cm.gray)

Out[13]:

<matplotlib.image.AxesImage at 0x104f7bc50>

In [14]:

plt.imshow(noise(), cmap=plt.cm.gray)

Out[14]:

<matplotlib.image.AxesImage at 0x104ff2d90>

Using ARMA processes¶

Now, we define the ARMA process as an averaging process with a certain time constant $\tau=30.$ (in frames).

In [15]:

def ARMA(image, tau=30.):
    image = (1 - 1/tau)* image + 1/tau * noise()
    return image

initializing

In [17]:

image = ARMA(lena)
plt.imshow(image, cmap=plt.cm.gray)

Out[17]:

<matplotlib.image.AxesImage at 0x106d75990>

In [18]:

for _ in range(1000): image = ARMA(image)
plt.imshow(image, cmap=plt.cm.gray)

Out[18]:

<matplotlib.image.AxesImage at 0x1076aff90>

In [19]:

for _ in range(1000): image = ARMA(image)
plt.imshow(image, cmap=plt.cm.gray)

Out[19]:

<matplotlib.image.AxesImage at 0x107d66490>

In [20]:

for _ in range(1000): image = ARMA(image)
plt.imshow(image, cmap=plt.cm.gray)

Out[20]:

<matplotlib.image.AxesImage at 0x107d98950>

In [21]:

N_frame = 1024
z = np.zeros((lena.shape[0], lena.shape[1], N_frame))
z[:, :, 0] = image
for i_frame in range(1, N_frame): 
    z[:, :, i_frame] = ARMA(z[:, :, i_frame-1])
mc.anim_save(.5 + .5*z, filename='results/arma')

calculating  99% |############################################# | ETA:  0:00:00

Saving sequence results/arma.mp4

calculating 100% |##############################################| Time: 0:00:58

In [22]:

show_video(filename='arma')

In []:

Testing basic functions of Motion Clouds

Motion Clouds: raw principles¶

Using Fourier ("official Motion Clouds")¶

Using mixtures of images¶

Using ARMA processes¶